Sealand: The fusion and symbiosis of cloud and AI, long-term benefits from computing power leasing

Sealand released a research report stating that Meta's transformation to cloud computing is mainly due to considerations of its own business model, and the long-term growth of AI computing power is still expected to be sustained. It is expected that the supply and demand imbalance of AI computing power will intensify in China, and the long-term growth logic of the computing power rental industry is relatively certain. It has entered a period of dividend release in performance, maintaining a "recommended" rating for the computer industry. Sealand's main points are as follows: AI and cloud integration, cloud business is an optimized complement to Meta's business model According to GMTEight, Meta is developing a new cloud computing infrastructure business plan, which will sell access to AI computing infrastructure and exclusive AI large model resources deployed by Meta on a large scale. The firm believes that renting computing power by Meta is in line with the main laws of development of the AI industry, where AI and cloud integration are symbiotic. From 2024 to 2026, large Internet companies in North America continue to increase their AI capital spending, competing in large model competitions. In 2025, the total capital spending of the top five Internet companies, Amazon.com, Inc., Alphabet Inc. Class C, Microsoft Corporation, Meta, and Oracle, reached 449 billion U.S. dollars, a year-on-year increase of 72%. The firm expects that the total spending of these five major companies will reach 814.5 billion U.S. dollars in 2026, an increase of 81% year-on-year. Cloud computing business is an effective supplement to asset utilization for AI companies with large-scale hardware procurement. The firm predicts that Meta's cumulative capital spending from 2024 to 2026 will be 246.4 billion U.S. dollars, and the launch of cloud business will help improve its AI computing power asset utilization, with cloud revenue benefiting the core business of the company and promoting future capital spending growth for Meta. NVIDIA Corporation expects the overall annual AI industry spending to reach 30 trillion to 40 trillion U.S. dollars by the end of 2030, with the cumulative revenue of the Blackwell and Vera Rubin dual products expected to reach 1 trillion U.S. dollars from 2025 to 2027. According to Fast Technology, Taiwan Semiconductor Manufacturing Co., Ltd. Sponsored ADR is expected to have a capital expenditure of 52-56 billion U.S. dollars in 2026, leaning towards the upper limit, stating that "AI demand is very strong," and that the world is currently in the "AI trend," with capital expenditure for the next three years expected to be significantly higher than the previous three years. The AI supply and demand in the Chinese market may be more tense, and the performance advantage of B300 servers is significant Domestic CSPs are expected to continue to increase their capital spending on computing power. In 2025, TENCENT and Alibaba Group Holding Limited Sponsored ADR's capital expenditures were 79.2 billion and 123.1 billion respectively, representing year-on-year growth of 3% and 72%. Since 2026, domestic CSPs have increased additional spending. Alibaba Group Holding Limited Sponsored ADR is considering increasing its investment in AI infrastructure and cloud computing from 380 billion yuan in the next three years (2025-2027) to 480 billion yuan. Bytedance has raised its AI capital spending plan for 2026 to over 20 billion yuan. The firm predicts that due to the rapid development of domestic computing power and computing power rental industries, domestic CSPs are expected to quickly increase their infrastructure supply in AI. In 2026, TENCENT, Alibaba Group Holding Limited Sponsored ADR, and Bytedance's capital expenditures are expected to reach 131.9 billion, 180 billion, and 285.7 billion respectively, representing year-on-year increases of 67%, 46%, and 90%, and the potential growth of domestic CSPs in AI computing power is considerable. The transaction price of NVIDIA B300 AI servers in mainland China has risen. Fast Technology reported on May 1st that the transaction price of NVIDIA B300 AI servers in the Chinese mainland market has risen to about 7 million yuan per unit, equivalent to about 1 million U.S. dollars, nearly doubling from the end of last year. The B300 server has shown a significant improvement in inference performance compared to the H200 and its price is expected to further rise. Domestic computing power rental companies are increasing their spending, with optimistic expectations for future profit growth In 2026, domestic computing power rental companies are expected to gradually enter a period of high growth in performance: 1) Sharetronic Data Technology: achieved operating income of 6.085 billion yuan in 2026 Q1, a year-on-year increase of 192.90%; net profit attributable to owners of the company reached 750 million yuan, a year-on-year increase of 343.45%, and a quarterly increase of 60%. 2) Jiangsu Lettall Electronic: achieved operating income of 997 million yuan in 2026 Q1, a year-on-year increase of 41.61%; net profit attributable to owners of the company was 271 million yuan, a significant increase of 821.08% year-on-year and 253.60% quarter-on-quarter. 3) Glory View Technology: as of 2026 Q1, the total assets of the company have increased significantly to 12.945 billion yuan, a 143% increase from the end of 2025, with construction in progress amounting to 1.515 billion yuan, a more than 17.41-fold increase year-on-year. Representative companies in the computing rental industry are increasing their financing and computing power procurement efforts to accelerate their expansion. They are relying on large credit lines and hardware procurement channels to accelerate the layout of computing power clusters. The firm believes that the current rapid expansion of the domestic computing rental industry, with companies leveraging large-scale computing power reserves and long-term customer contracts, has achieved profit nonlinear growth multiple times higher than revenue growth in 2026 Q1. Their supply chain value is expected to further expand in the future. Related companies: 1) CPU: Hygon Information Technology, China Greatwall Technology Group (Feiteng Information); 2) Connectivity: Montage Technology, Shengketong Communications, Vantone Neo Development Group, Ruijie Networks, Sichuan Huafeng Technology, Guizhou Space Appliance, Hgtech; 3) AI chips: Cambricon, BIREN TECH, Verisilicon Microelectronics (Shanghai) Co., Ltd., Muxi Stock, ILUVATAR COREX; 4) Computing power rental: Sharetronic Data Technology, Glory View Technology, Jiangsu Lettall Electronic, Guangzhou Sie Consulting, Maxvision Technology Corp., Jwipc Technology, Range Intelligent Computing Technology Group, RunJian Co., Ltd .; 2) Super node OEMs: Inspur Electronic Information Industry, Dawning Information Industry, Unisplendour Corporation, Huaqin Co., Ltd., Foxconn Industrial Internet. Risk warning: Escalation of US-China tensions, downstream demand affected by macroeconomic factors, AI inference demand falling short of expectations, intensifying market competition, supply chain risks, etc.